OmniSequence: Transfinite 2.0.10

English

Performance

• Optimized external pattern access for the Sequence Array. Ordinary items and unsupported patterns are rejected before expensive structure checks, reducing the overhead of repeated insertion attempts from Storage Buses and automation.
• Reduced repeated quantum-crystal scans. Reuse the positions of all 14 crystals for each structure orientation and combine presence and data-ownership checks into one live pass. Crystal removal, replacement and pattern changes within the same tick are still detected; stale caches cannot restore patterns that were moved out.
• Optimized moving Singularity Hub structures. Reuse collision geometry and coordinate transforms, reducing repeated obstacle calculations and unnecessary fluid scans while retaining precise collisions, passenger carriage, docking, saved-state migration and ahead-of-motion obstacle checks.
• Improved molecular-device batch dispatch. Reduce repeated provider probes and improve adaptive-batch recovery after temporary rejection. Ordinary and reusable-tool paths share per-tick batch accounting and correctly release rejected reservations. Existing logical parallelism limits are not reduced.

Crafting CPUs and ECO Dispatch

• Improved dispatch for the Omni-Computation Core and Transfinite Compute Nexus. Virtual CPU lanes fairly share dispatch budgets, with consistent work, capacity and fallback accounting to improve behavior when large orders and concurrent jobs compete.
• Improved ECO plan integration. Strengthen the connection between Omni CPUs and exact-count, phased and cyclic plans, as well as large-order parent jobs. Preserve job UUIDs, outstanding outputs, phase progress and reload state. Parent orders advance only when outputs are actually delivered, never crediting undelivered work as complete.
• Fixed ECO selecting the wrong dispatch path for the Matter Fabrication Well. The well no longer enters the generic AE2 processing-pattern ramp first. It uses its existing native batch transaction instead, preventing small incremental dispatches from blocking the native batch interface.
• Completed native ECO exact-batch preparation. Transfer ownership of the complete input vector, accepting a whole batch that fits the current long-sized capacity while leaving the excess in the CPU. Large counts are not truncated, and smaller input vectors cannot be substituted to fabricate capacity.
• Strengthened resource conservation during rejection, cancellation and reload. Missing catalysts or dispatch power do not consume materials. Cancellation refunds only the queued and active inputs owned by the corresponding job UUID, without affecting other jobs using the same pattern. Reload and repeated cancellation do not duplicate refunds.
• Improved fabrication progress handling. Preserve and rescale existing progress when processing duration changes. Instant-batch outputs wait for the corresponding CPU accounting before returning to the network, preventing premature feedback.

Display and Compatibility Fixes

• Fixed collision collection for moving structures with Lithium installed. Add the moving Singularity Hub's precise collision surfaces to Lithium's deferred collision query while retaining the vanilla query path. This compatibility path is not loaded when Lithium is absent.

中文

性能优化

• 优化构序阵列的外部样板访问。 普通物品及不支持的样板会在昂贵的结构检查前被拒绝，减少存储总线和自动化反复尝试插入时的额外开销。
• 减少量子水晶重复扫描。 按结构朝向复用 14 个水晶的位置，并将存在性与数据所有权检查合并为一轮实时校验；同一 tick 内拆除、更换水晶或修改样板仍能被检测，不会用旧缓存恢复已移走的样板。
• 优化奇点天枢运动结构。 复用碰撞几何与坐标变换，减少障碍检测中的重复计算和不必要的流体扫描；保留精确碰撞、乘载、停靠、存档迁移及前方障碍检测。
• 改进分子设备批量派发。 减少重复供应器探测，改善自适应批次在临时拒绝后的恢复；普通与可复用工具路径共用单 tick 批量记账，并正确退回失败预留。原有逻辑并行上限不降低。

合成 CPU 与 ECO 调度

• 完善万物演算核心和超限算枢的派发调度。 多个虚拟 CPU 通道公平共享派发额度，统一工作量、容量与回退路径的记账，改善大订单长期占用和多任务竞争时的行为。
• 完善 ECO 计划接入。 加强精确数量、阶段计划、循环计划及大单父订单与 Omni CPU 的衔接，保留任务 UUID、等待产物、阶段进度及重载状态；父订单按实际交付推进，不把尚未交付的部分提前计为完成。
• 修复 ECO 向物质构筑井投递时选错派发路径的问题。 构筑井不再先进入通用 AE2 处理样板的渐进小批次路径，而是使用已有的原生批量事务，避免批量接口被小批次调度挡住。
• 补齐 ECO 原生精确批量准备。 按完整材料向量接管输入，单批接纳当前 long 容量能够表达的完整数量，超出部分继续留在 CPU；不截断大数，也不通过缩小输入伪造容量。
• 加强拒绝、取消与重载的资源守恒。 缺少催化剂或派发电力时不扣料；取消任务只退还对应 UUID 的排队和执行中原料，不影响使用同一样板的其他任务；重载和重复取消不会重复退款。
• 改进构筑井加工状态。 加工耗时变化时保留并换算已有进度；即时批次的产物等待 CPU 完成相应记账后再回传，避免提前回流。

显示与兼容修复

• 修复 Lithium 环境下运动结构的碰撞收集。 将天枢运动结构的精确碰撞面补入 Lithium 的延迟碰撞查询，保留原版查询路径；未安装 Lithium 时不加载该兼容路径。